2 min readaiask-aiplatformannouncement

Ask AI, Your Choice of Model: Auto, DeepSeek and Sonnet 5

TestRelic Team

Every Ask AI answer used to come from the same model. Now you choose, the answer tells you which model wrote it, and an admin can settle the question for the whole organization.

Pick the engine

The Ask AI composer has a model picker with five choices: Auto, DeepSeek V4 Pro, DeepSeek Flash, Sonnet 4.6 and Sonnet 5. Each person's choice is saved.

Auto starts on DeepSeek Flash and brings in Sonnet 4.6 when a turn needs it:

  • Turns with images or web search go to Sonnet, because DeepSeek can't read images or search the web.
  • If DeepSeek fails before an answer starts, including when it can't be reached at all, the turn switches to Sonnet once rather than erroring.

Every answer names its model

Each answer carries an "Answered by" receipt naming the exact model, and says so when Auto switched engines. When you're comparing two answers, or deciding whether a report is worth regenerating on a bigger model, you know what you're looking at.

One policy for the whole organization

Admins can set Ask AI engine in Organization settings. Leave it on Members choose, and everyone picks in the composer. Or pin one engine (Auto, V4 Pro, Flash, Sonnet 4.6 or Sonnet 5) for everyone.

Pinning is a deliberate trade:

  • A pinned engine has no automatic failover. If it's down, turns fail rather than quietly moving to another model.
  • Images and web search still go to Sonnet, even with a DeepSeek engine pinned.
  • Where your conversations are processed follows the engine. DeepSeek models run on DeepSeek's servers and cost substantially less; Sonnet runs on Anthropic's. An organization that wants every turn on Anthropic can pin a Sonnet engine.

Follow Studio and CLI conversations from the web

The Ask AI sidebar now opens with a Chats / Remote switcher. Chats is your web conversations, as before. Remote lists Ask AI conversations running in Studio or the CLI, live ones first, and lets you watch a turn as it streams, including the artifacts it produces along the way. There's nothing to set up: conversations from Studio and the CLI appear on their own.

Get started

Open Ask AI and choose a model from the picker in the composer. If you're an admin, set the organization's policy under Organization settings → Ask AI engine. Ask AI is part of the Growth plan and above; see Plans & Billing.